Neroni reframes creativity's unit of explanation from isolated factors to cross-level dynamic configurations, offering four mechanisms—constraint, affordance, regulation, feedback/selection—and testable propositions

The paper shifts the unit of explanation from isolated determinants to the changing configuration of relationships across levels, arguing that the same factor—such as cognitive flexibility, intrinsic motivation, social support, domain expertise, or technological assistance—may support creativity under some conditions while having little effect or even becoming constraining under others. Relative to Csikszentmihalyi's systems model, Glăveanu's sociocultural approach, and the Person × Task × Situation framework, the paper extends the relational view across levels: psychological and sociocultural processes are themselves shaped by embodied and neurobiological conditions, developmental trajectories, and increasingly by technological systems. This is a theoretical Perspective article; its argument rests on integrating and reinterpreting literature from neuroscience, psychology, sociocultural research, education, organizational studies, and generative AI rather than on new empirical data, and the author states the aim is not a comprehensive review of creativity research.

The paper proposes four mechanisms—constraint, affordance, regulation, and feedback/selection—to describe how biological, psychological, social, cultural, and technological processes influence one another over time, and derives four testable propositions: configurational dependence, temporal specificity, cross-level divergence, and developmental reconfiguration. The author argues these propositions distinguish the framework from the general claim that 'everything interacts': if major creativity-related factors consistently showed the same effects across tasks, phases, levels of analysis, and developmental conditions, this would challenge the framework, whereas systematic moderation, changes over time, cross-level trade-offs, and developmental differences would support a configurational account. The four propositions are stated as conditional predictions that the author describes as empirically testable, but the text reports no new data testing them.

The paper uses human–AI creativity as an illustration of a configurational rather than additive phenomenon: in Chen and Chan's advertising task study, using a large language model as a 'sounding board' improved outcomes for non-experts, whereas using it as a 'ghostwriter' provided no comparable advantage and was detrimental to expert users and could produce an anchoring effect; McGuire et al. found participants produced less creative poetry when given an AI-generated poem to edit, but this disadvantage disappeared when the system was redesigned for iterative co-creation, which also preserved higher creative self-efficacy. These results are used to support the claim that the same technological capability functions differently depending on user expertise and the role assigned to AI in the task, shifting the question from whether AI enhances creativity to under which human–task–technology configurations it does so. Evidence comes from the specific experimental studies cited and from Vaccaro et al.'s systematic review and meta-analysis, which found substantial variation in human–AI performance depending on the relative capabilities of humans and AI and how their contributions were combined, though those findings extend beyond creativity specifically.

The paper argues that contradictory findings should be treated as theoretically informative and calls for greater conceptual and measurement specificity: divergent-thinking performance, creative potential, everyday creative behavior, domain-specific products, and socially recognized creative achievement are related but distinct phenomena operating over different timescales. The author accordingly reframes the empirical question from 'Does X increase creativity?' to 'For whom does X matter, at which stage of creative activity, in combination with which other resources or constraints, at which level of analysis, and over what period of time?' This claim builds on a synthesis of heterogeneity in existing literature, including critiques of generalizing default mode network findings to creativity as a whole, arguments against treating creativity as a single cognitive faculty, and meta-analytic findings on gender differences; the text offers no new psychometric validation.

The framework is aimed at researchers studying creativity across levels and at practitioners designing and evaluating education, organizational, and human–AI collaboration settings. It applies under the premise that creativity is a changing multilevel configuration rather than a stable individual property; it therefore advocates selecting measures according to the level, phase, and timescale of the phenomenon studied, and using longitudinal, intensive repeated-measures, experience-sampling, multilevel modeling, social-network, and digital trace methods. The author also notes that the four mechanisms are neither exhaustive nor mutually exclusive, and that this does not imply all levels are equally important in every creative act.

It is worth noting that this is a theoretical Perspective, the four propositions await independent empirical testing, and the text provides no unified effect sizes or sample-level synthesis. The cited human–AI findings come from specific tasks (creative advertising, poetry writing) and specific system configurations, so generalization to other domains and longer timescales remains an open question. In addition, the available text is an incomplete version with gaps in the reference list and some cited authors and years appearing as bracketed placeholders in the body, so not all citation details can be verified; readers needing to trace a specific piece of evidence should consult the complete published version.